When and How Much to Imagine: Adaptive Test-Time Scaling with World Models for Visual Spatial Reasoning

TL;DR

This paper introduces AVIC, an adaptive framework that controls when and how much visual imagination is used during spatial reasoning tasks, improving efficiency and accuracy by selectively invoking world models.

Contribution

It presents a novel adaptive test-time approach that explicitly reasons about the necessity of visual imagination, optimizing its use for better spatial reasoning performance.

Findings

Selective imagination improves reasoning accuracy.

Adaptive control reduces computational cost.

Proper management of imagination prevents performance degradation.

Abstract

Despite rapid progress in Multimodal Large Language Models (MLLMs), visual spatial reasoning remains unreliable when correct answers depend on how a scene would appear under unseen or alternative viewpoints. Recent work addresses this by augmenting reasoning with world models for visual imagination, but questions such as when imagination is actually necessary, how much of it is beneficial, and when it becomes harmful, remain poorly understood. In practice, indiscriminate imagination can increase computation and even degrade performance by introducing misleading evidence. In this work, we present an in-depth analysis of test-time visual imagination as a controllable resource for spatial reasoning. We study when static visual evidence is sufficient, when imagination improves reasoning, and how excessive or unnecessary imagination affects accuracy and efficiency. To support this analysis, we introduce AVIC, an adaptive test-time framework with world models that explicitly reasons about the sufficiency of current visual evidence before selectively invoking and scaling visual imagination. Across spatial reasoning benchmarks (SAT, MMSI) and an embodied navigation benchmark (R2R), our results reveal clear scenarios where imagination is critical, marginal, or detrimental, and show that selective control can match or outperform fixed imagination strategies with substantially fewer world-model calls and language tokens. Overall, our findings highlight the importance of analyzing and controlling test-time imagination for efficient and reliable spatial reasoning.